Intelligent Ubiquitous Sensor Network for ... - Semantic Scholar
Intelligent Ubiquitous Sensor Network for Agricultural and Livestock Farms Junghoon Lee1 , Hye-Jin Kim1 , Gyung-Leen Park1, Ho-Young Kwak2 , and Cheol Min Kim3 1
Dept. of Computer Science and Statistics 2 Dept. of Computer Engineering 3 Dept. of Computer Education Jeju National University, 690-756, Jeju-Do, Republic of Korea {jhlee,hjkim82,glpark,kwak,cmkim}@jejunu.ac.kr
Abstract. This paper designs and implements an intelligent ubiquitous sensor network architecture for agricultural and livestock farms which embrace a variety of sensors and create a great volume of sensor data records. For the sake of efficiently and accurately detecting the specific events out of the great amount of sensor data which may include not just erroneous terms but also correlative attributes, the middleware module embeds an empirical event patterns and knowledge description. For the filtered data, data mining module opens an interface to define the relationship between the environmental aspect and facility control equipments, set the control action trigger condition, and integrate new event detection logic. Finally, the remote user interface for monitoring and control is implemented by on Microsoft Windows, Web, and mobile device applications. Keywords: Ubiquitous sensor network, middleware, rule-based data processing, event detection, control box interface.
1
Introduction
Nowadays, wireless sensor networks have been successfully applied to environmental and wildlife habitat monitoring [1], while its intelligent and efficient management improves productivity and revenue of the agricultural and livestock farms [2]. Sensor data, inherently quite different from the traditional data records, are created in the form of a real-time, continuous, ordered sequence of sensor readings. Here, the temporal order can be decided either implicitly by arrival time or explicitly by timestamp, so a data stream is defined as a continuous sequence of tuples. Structure of data items in a data stream can change in time. Moreover, many data streams can include the spatial tag not just the temporal order, possibly hosting a geographic application on the sensor network.
This research was supported by the MKE (The Ministry of Knowledge Economy), through the project of Region technical renovation, Republic of Korea.
Y. Xiang et al. (Eds.): ICA3PP 2011 Workshops, Part II, LNCS 7017, pp. 196–204, 2011. c Springer-Verlag Berlin Heidelberg 2011
Intelligent Ubiquitous Sensor Network for Agricultural and Livestock Farms
197
A sensor network can be viewed as a large database system which responds to the query issued from various applications [3,4]. For example, the SyncQuery language that expresses composable queries over streams, pointing out that composition of queries, and hence supporting views is not possible in the append-only stream model [5]. This language employs the tagged stream model in which a data stream is treated as a sequence of modifications over the given relation. Particularly, the sliding-window approach is generalized by introducing the synchronization principle that empowers SyncSQL with a formal mechanism to express queries with arbitrary refresh condition. Besides, this work includes an algebraic framework for SyncSQL queries, couple of equivalences and transformation rules, and a query-matching algorithm. The main task of ubiquitous sensor networks, or USN in short, is monitoring sensor values, deciding the control actions, and triggering appropriate actuators [6]. For example, if the current lightness drops below the permissible level, USN can turn on the light in the green house. Moreover, if the current CO2 level is higher than a specific bound, a ventilator is activated to refresh the air. To this end, a lot of sensors are installed over the wide target area and each of them reports its sensor values to the controller, creating a tremendous amount of data records. USN must handle the large volume of sensor records and analyze them. Here, more than one sensor records are correlated as they capture the same event, and the records have sequential or spatial correlation. Moreover, sensor values can have garbage and measurement errors. The instability of wireless networks also can jeopardize the correct analysis. Wrong reaction stemmed from wrong data analysis can burn actuator motors, waste power, and lead to many hazardous problems. In this regard, this paper is to design and implement a USN architecture for agricultural and livestock farms, aiming at efficiently and accurately handling great volume of sensor data obtained from a variety of sensor devices and generating a correct control action. Our implementation focuses on the data processing middleware that interacts with sensor nodes containing CO2 , temperature, humidity, lightness, and wind sensors. The system design opens an interface to define a rule to filter the raw data, correlate multiple streams, and decide the control action. Next, the remote user interface for monitoring and control USN is implemented by on Windows, Web, and mobile device applications. The rest of this paper is organized as follows: After issuing the problem in Section 1, Section 2 describes the background and related work, focusing on the target USN architecture. Section 3 describes raw data processing and middleware processing of the proposed system, respectively. Section 4 presents the user interface implementation details. Finally, Section 5 concludes this paper with a brief introduction of future work.
2
Background and Related Work
Under the research and technical project named Development of convergence techniques for agriculture, fisheries, and livestock industries based on the
198
J. Lee et al.
ubiquitous sensor networks, our project team has designed and developed an intelligent USN framework [7]. This framework provides an efficient and seamless runtime environment for a variety of monitor-and-control applications on sensor networks. The sensor node, built on the Berkeley mote platform, comprises sensors, microprocessor, radio transceiver, and battery [8]. Over the sensor network mainly exploiting the Zigbee technology, composite sensors detect events such as body heat change of a livestock via the biosensors attached to it as well as humidity, CO2 , and NH3 level via the environmental sensors. Each node runs the IP-USN protocol and implements corresponding routing schemes [9]. The sensor network and the global network, namely, the Internet, are connected through the USN gateway. At this stage, the system is to integrate a remote control model to provide remote irrigation and the activation of heater or fan.
Fig. 1. Agricultural USN framework
Our previous work has designed an intelligent data processing framework in ubiquitous sensor networks, implementing its prototype [7]. Much focus is put on how to handle the sensor data stream as well as the interoperability between the low-level sensor data and application clients. This work first designs systematic middleware which mitigates the interaction between the application layer and low-level sensors, for the sake of analyzing a great volume of sensor data by filtering and integrating to create value-added context information. Then, an agent-based architecture is proposed for real-time data distribution to forward a specific event to the appropriate application, which is registered in the directory service via the open interface. The prototype implementation demonstrates that this framework can not only host a sophisticated application on USN and but also autonomously evolve to new middleware, taking advantages of promising technologies such as software agents, XML, and the like. Particularly, cloud
Intelligent Ubiquitous Sensor Network for Agricultural and Livestock Farms
199
computing can provide the high-speed data processing framework for sensor streams [10]. It must be mentioned that XML data stream processing is also of interest, as XML becomes a common part of information systems, including RFID (Radio Frequency IDentifier), ad-hoc sensor data collection, network traffic management, and so-called service-oriented architecture [11]. Generally, XML streams are created as a second-hand product obtained from information exchange in XML systems, rather than from the raw sensor values. XML data streams can be viewed as a sequence of XML documents, and each data item in the stream is a valid standalone XML document, which is independent of other items in the stream. Moreover, queries on data stream can support data mining and filtering. While the first evaluates queries that span over a long time period, processing a great deal of time-sequenced data, the second takes the data items from the stream matching the filtering condition. Anyway, processing XML has an attractive real-world motivation and our system will also take advantage of XML technologies for interactions between data processing modules.
3 3.1
Intelligent USN Architecture Raw Data Processing
Each sensor output must be converted to our daily-life values. First, the sensor board consistently supplies 2500 mV to the soil humidity sensor device, which will generate the voltage value of 250 through 1000 mV . Here, 250 mV corresponds to 0 % humidity while 1000 mV to 100 % humidity. Next, pyranometer sensors are used to measure the solar radiation flux density on a planar surface, generally in watts per meter square. According to the sensor device specification, 220 mV is detected on full sunlight, namely, 1100 W m−2 . Hence, by Eq. (1), we can obtain the solar radiation value. Sr = So × Cv = 200mV × 5.0W −2 /mV = 1100W −2,
(1)
where Sr is solar radiation, So is the sensor output, and Cv is the conversion factor having 5 W m−2 /mV . Anometer sensors, commonly used in a weather station instrument, measures wind speed and direction. The device calculates the wind direction based on the probed voltage values measured from different angles. The relationship between the difference and the phase angle is provided as shown in Figure 2 and the corresponding measurement value is estimated as in Eq. (2). Vout − 2431 (2) −6.8473 In addition, wind speed is measured by counting the number of rotations of a wind cup during the unit time. Namely, π Ws = × Nr , (3) t θ[deg] =
200
J. Lee et al.
Output voltage (mV)
2500 "OutputVoltage" 2000 1500 1000 500 0 0
45
90
135
180 225 Degrees
270
315
360
Fig. 2. Phase angle and wind direction
where Ws is the wind speed estimation and Nr is the number of rotations during time interval t. 3.2
Middleware Layer Processing
Middleware works between the sensor interface and high-end data analyzers as shown in Figure 3. To begin with, as the collected data may have erroneous readings and garbage values, the middleware is required to check the validity range of the collected data first and prevent multiple reactions to a single event. To this end, the duplicate detector module checks the event length and value changes for the real-time sensor data, storing the filtered event in the database. The controller logic can define the control target and reaction range to activate the predefined control action to a specific event. Now, from the database, the meaningful context is extracted by the sophisticated classification and time series analysis for the event-level sequences. The series of event patterns and interpreted knowledge are embedded in the analysis module to recognize abnormal conditions instantly. Moreover, our system opens an interface to define the relationship between the environmental aspect and facility control equipments, set the control action trigger condition, and integrate new event detection logic. The inference engine defines a set of rules to detect events. To define a rule, each sensor and node is assigned a unique identifier, while max(), min(), average(), count(), and run() functions are provided for better event specification. Using this, we can specify several rules, for example, report an event when average temperature of node 123 is higher than 35, or turn on all fans installed in sensor node 452. Based on this rule-base, the middleware checks the validity of the sensor data and requests the retransmission if it has an error term. After calculating the difference from the previous sensor reading, the middleware detects an abnormal condition based on the empirically obtained event patterns and knowledge. This procedure is illustrated in Figure 4.
Intelligent Ubiquitous Sensor Network for Agricultural and Livestock Farms
Data logging mgmt Data management
201
Sensor management Sensor group mgmt dynamic event
Query processing
MSMQ (MicroSoft Message Queue) Data conversion
excel extension, XML exchange
Error processing
exception check, error policy
Duplication processing
duplicate sensor data check
Event processing
event detection logic
Sensor interface Fig. 3. Middleware architecture Rule
Event detection Rule Sensor info Database manager Rule
Data mining
Sensor value Log
Rule−based value processing
Response & event Request
Actuator control process
Data manager
Sensor value
Data source (Sensor nodes)
Fig. 4. Event logic
4
User Interface and Control Action
Remote monitor-and-control keeps track of the environmental sensor values on temperature, humidity, lightness, and CO2 . It can turn on or off the power switch to each actuator device. For example, the temperature monitor tracks the current temperature of a specific position selected via the geographic map. According to the initiation command, the server module begins to collect and store sensor readings. During the lifetime of this operation, the event detection is carried out based on the criteria specified in the query. The client also retrieves the current temperature value to monitor the up-to-date temperature change. Figure 5 show the user interface implemented in this application. First of all, it displays the
202
J. Lee et al.
(a) Normal status
(b) Abnormal status detection Fig. 5. User interface
Fig. 6. Control box interface
map, location of sensors along with the current status of them. In addition, the series of sensor values are scrolled in the listbox, while a graph is created to plot the temperature change. Figure 5(a) indicates the normal status where no sensor value deviates from the given bound, and all nodes are marked blue. Whereas,
Intelligent Ubiquitous Sensor Network for Agricultural and Livestock Farms
203
in Figure 5(b), one sensor node detects the value out of the normal range and this node turns red. In addition, a remote control interface is implemented in an embedded control box and a u-Multi smart mote. First, the embedded control box application interacts with the control system via TCP/IP. As a Microsoft Windows application, it sends a control command such as current status retrieval and specific control action trigger, as shown in Figure 6. In response to this command, the sensor network sends an acknowledgment back to the controller box. In addition, the sensor network can automatically notify the event of approaching the final permissible borderline of current sensor reading. This control box application is also implemented as a web program. Second, u-Multi smart mote is functionally similar to the control box except that its communication interface is SMS (Short Message Service) instead of TCP/IP and it is developed on a smaller user display. One of the most critical events is power breakage and the battery can survive tens of minutes to allow a failure recovery procedure, including notification to human managers.
5
Concluding Remarks
In this paper, we have designed and developed an intelligent ubiquitous sensor network targeting at agricultural and livestock farms which have a variety of sensors and create a great volume of sensor data records during the monitoring phase. For the sake of efficiently and accurately detecting the specific events out of the great number of sensor data which may include erroneous terms, correlative components, the middleware module embeds an empirical event patterns and knowledge description. It also interprets sensor-specific data to the actual values. For the filtered data, data mining module opens an interface to define the relationship between the environmental aspect and facility control equipments, set the control action trigger condition, and integrate new event detection logic. Finally, the remote user interface for monitoring and control USN is implemented in Windows, Web and mobile device applications. As future work, we are planning to design an advanced data inference engine for management information as well as sensor data [13]. The sophisticated data analysis will create a new type of management messages and those messages will make USN more intelligent.
References 1. Golab, L., Oszu, M.: Issues in Data Stream Management. ACM SIGMOD Record 32, 5–14 (2003) 2. Lee, J., Park, G., Kim, H., Kim, C., Kwak, H., Lee, S., Lee, S.: Intelligent Management Message Routing in Ubiquitous Sensor Networks. In: Int’l. Conference on Computational Collective Intelligence-Technologies and Applications (2011) 3. Madden, S., Franklin, M., Hellerstein, J., Hong, W.: TinyDB: an Acquisitional Query Processing System for Sensor Networks. ACM Transactions on Database Systems 30 (2005)
204
J. Lee et al.
4. Madden, S., Franklin, M.: Fjording the Stream: An Architecture for Queries over Streaming Sensor Data. In: Proc. of the 2002 Intl. Conf. on Data Engineering (2002) 5. Ghanem, T., Elmagarmid, A., Larson, P., Aref, W.: Supporting Views in Data Stream Management Systems. ACM Transactions on Database Systems 35(1) (2010) 6. Culler, D., Estrin, D., Srivastava, M.: Overview of Sensor Networks. IEEE Computer 37, 41–49 (2004) 7. Lee, J., Park, G., Kwak, H., Kim, C.: Efficient and Extensible Data Processing Framework in Ubiquitous Sensor Networks. In: International Conference on Intelligent Control Systems Engineering, pp. 324–327 (2011) 8. http://www.tinyos.net 9. Cuevas, A., Urue˜ na, M., Laube, A., Gomez, L.: LWESP: Light-Weight Exterior Sensornet Protocol. In: IEEE International Conference in Computer and Communications (2009) 10. Kang, M., Kang, D., Crago, S., Park, G., Lee, J.: Design and Development of a Run-time Monitor for Multi-Core Architectures in Cloud Computing. Sensors 11, 3595–3610 (2011) 11. Ulrych, J.: Processing XML data streams: A survey. In: WDS Proc. Contributed Papers, pp. 218–223 (2008) 12. Madden, S., Franklin, M., Hellerstein, J., Hong, W.: The Design of an Acquisitional Query Processor for Sensor Networks. In: ACM SINGMOD (2003) 13. Woo, H., Mok, A.: Real-Time Monitoring of Uncertain Data Streams Using Probabilistic Similarity. In: Proc. of IEEE Real-Time Systems Symposium, pp. 288–300 (2007)
Dept. of Computer Science and Statistics 2 Dept. of Computer Engineering 3 Dept. of Computer Education Jeju National University, 690-756, Jeju-Do, Republic of Korea {jhlee,hjkim82,glpark,kwak,cmkim}@jejunu.ac.kr
Abstract. This paper designs and implements an intelligent ubiquitous sensor network architecture for agricultural and livestock farms which embrace a variety of sensors and create a great volume of sensor data records. For the sake of efficiently and accurately detecting the specific events out of the great amount of sensor data which may include not just erroneous terms but also correlative attributes, the middleware module embeds an empirical event patterns and knowledge description. For the filtered data, data mining module opens an interface to define the relationship between the environmental aspect and facility control equipments, set the control action trigger condition, and integrate new event detection logic. Finally, the remote user interface for monitoring and control is implemented by on Microsoft Windows, Web, and mobile device applications. Keywords: Ubiquitous sensor network, middleware, rule-based data processing, event detection, control box interface.
1
Introduction
Nowadays, wireless sensor networks have been successfully applied to environmental and wildlife habitat monitoring [1], while its intelligent and efficient management improves productivity and revenue of the agricultural and livestock farms [2]. Sensor data, inherently quite different from the traditional data records, are created in the form of a real-time, continuous, ordered sequence of sensor readings. Here, the temporal order can be decided either implicitly by arrival time or explicitly by timestamp, so a data stream is defined as a continuous sequence of tuples. Structure of data items in a data stream can change in time. Moreover, many data streams can include the spatial tag not just the temporal order, possibly hosting a geographic application on the sensor network.
This research was supported by the MKE (The Ministry of Knowledge Economy), through the project of Region technical renovation, Republic of Korea.
Y. Xiang et al. (Eds.): ICA3PP 2011 Workshops, Part II, LNCS 7017, pp. 196–204, 2011. c Springer-Verlag Berlin Heidelberg 2011
Intelligent Ubiquitous Sensor Network for Agricultural and Livestock Farms
197
A sensor network can be viewed as a large database system which responds to the query issued from various applications [3,4]. For example, the SyncQuery language that expresses composable queries over streams, pointing out that composition of queries, and hence supporting views is not possible in the append-only stream model [5]. This language employs the tagged stream model in which a data stream is treated as a sequence of modifications over the given relation. Particularly, the sliding-window approach is generalized by introducing the synchronization principle that empowers SyncSQL with a formal mechanism to express queries with arbitrary refresh condition. Besides, this work includes an algebraic framework for SyncSQL queries, couple of equivalences and transformation rules, and a query-matching algorithm. The main task of ubiquitous sensor networks, or USN in short, is monitoring sensor values, deciding the control actions, and triggering appropriate actuators [6]. For example, if the current lightness drops below the permissible level, USN can turn on the light in the green house. Moreover, if the current CO2 level is higher than a specific bound, a ventilator is activated to refresh the air. To this end, a lot of sensors are installed over the wide target area and each of them reports its sensor values to the controller, creating a tremendous amount of data records. USN must handle the large volume of sensor records and analyze them. Here, more than one sensor records are correlated as they capture the same event, and the records have sequential or spatial correlation. Moreover, sensor values can have garbage and measurement errors. The instability of wireless networks also can jeopardize the correct analysis. Wrong reaction stemmed from wrong data analysis can burn actuator motors, waste power, and lead to many hazardous problems. In this regard, this paper is to design and implement a USN architecture for agricultural and livestock farms, aiming at efficiently and accurately handling great volume of sensor data obtained from a variety of sensor devices and generating a correct control action. Our implementation focuses on the data processing middleware that interacts with sensor nodes containing CO2 , temperature, humidity, lightness, and wind sensors. The system design opens an interface to define a rule to filter the raw data, correlate multiple streams, and decide the control action. Next, the remote user interface for monitoring and control USN is implemented by on Windows, Web, and mobile device applications. The rest of this paper is organized as follows: After issuing the problem in Section 1, Section 2 describes the background and related work, focusing on the target USN architecture. Section 3 describes raw data processing and middleware processing of the proposed system, respectively. Section 4 presents the user interface implementation details. Finally, Section 5 concludes this paper with a brief introduction of future work.
2
Background and Related Work
Under the research and technical project named Development of convergence techniques for agriculture, fisheries, and livestock industries based on the
198
J. Lee et al.
ubiquitous sensor networks, our project team has designed and developed an intelligent USN framework [7]. This framework provides an efficient and seamless runtime environment for a variety of monitor-and-control applications on sensor networks. The sensor node, built on the Berkeley mote platform, comprises sensors, microprocessor, radio transceiver, and battery [8]. Over the sensor network mainly exploiting the Zigbee technology, composite sensors detect events such as body heat change of a livestock via the biosensors attached to it as well as humidity, CO2 , and NH3 level via the environmental sensors. Each node runs the IP-USN protocol and implements corresponding routing schemes [9]. The sensor network and the global network, namely, the Internet, are connected through the USN gateway. At this stage, the system is to integrate a remote control model to provide remote irrigation and the activation of heater or fan.
Fig. 1. Agricultural USN framework
Our previous work has designed an intelligent data processing framework in ubiquitous sensor networks, implementing its prototype [7]. Much focus is put on how to handle the sensor data stream as well as the interoperability between the low-level sensor data and application clients. This work first designs systematic middleware which mitigates the interaction between the application layer and low-level sensors, for the sake of analyzing a great volume of sensor data by filtering and integrating to create value-added context information. Then, an agent-based architecture is proposed for real-time data distribution to forward a specific event to the appropriate application, which is registered in the directory service via the open interface. The prototype implementation demonstrates that this framework can not only host a sophisticated application on USN and but also autonomously evolve to new middleware, taking advantages of promising technologies such as software agents, XML, and the like. Particularly, cloud
Intelligent Ubiquitous Sensor Network for Agricultural and Livestock Farms
199
computing can provide the high-speed data processing framework for sensor streams [10]. It must be mentioned that XML data stream processing is also of interest, as XML becomes a common part of information systems, including RFID (Radio Frequency IDentifier), ad-hoc sensor data collection, network traffic management, and so-called service-oriented architecture [11]. Generally, XML streams are created as a second-hand product obtained from information exchange in XML systems, rather than from the raw sensor values. XML data streams can be viewed as a sequence of XML documents, and each data item in the stream is a valid standalone XML document, which is independent of other items in the stream. Moreover, queries on data stream can support data mining and filtering. While the first evaluates queries that span over a long time period, processing a great deal of time-sequenced data, the second takes the data items from the stream matching the filtering condition. Anyway, processing XML has an attractive real-world motivation and our system will also take advantage of XML technologies for interactions between data processing modules.
3 3.1
Intelligent USN Architecture Raw Data Processing
Each sensor output must be converted to our daily-life values. First, the sensor board consistently supplies 2500 mV to the soil humidity sensor device, which will generate the voltage value of 250 through 1000 mV . Here, 250 mV corresponds to 0 % humidity while 1000 mV to 100 % humidity. Next, pyranometer sensors are used to measure the solar radiation flux density on a planar surface, generally in watts per meter square. According to the sensor device specification, 220 mV is detected on full sunlight, namely, 1100 W m−2 . Hence, by Eq. (1), we can obtain the solar radiation value. Sr = So × Cv = 200mV × 5.0W −2 /mV = 1100W −2,
(1)
where Sr is solar radiation, So is the sensor output, and Cv is the conversion factor having 5 W m−2 /mV . Anometer sensors, commonly used in a weather station instrument, measures wind speed and direction. The device calculates the wind direction based on the probed voltage values measured from different angles. The relationship between the difference and the phase angle is provided as shown in Figure 2 and the corresponding measurement value is estimated as in Eq. (2). Vout − 2431 (2) −6.8473 In addition, wind speed is measured by counting the number of rotations of a wind cup during the unit time. Namely, π Ws = × Nr , (3) t θ[deg] =
200
J. Lee et al.
Output voltage (mV)
2500 "OutputVoltage" 2000 1500 1000 500 0 0
45
90
135
180 225 Degrees
270
315
360
Fig. 2. Phase angle and wind direction
where Ws is the wind speed estimation and Nr is the number of rotations during time interval t. 3.2
Middleware Layer Processing
Middleware works between the sensor interface and high-end data analyzers as shown in Figure 3. To begin with, as the collected data may have erroneous readings and garbage values, the middleware is required to check the validity range of the collected data first and prevent multiple reactions to a single event. To this end, the duplicate detector module checks the event length and value changes for the real-time sensor data, storing the filtered event in the database. The controller logic can define the control target and reaction range to activate the predefined control action to a specific event. Now, from the database, the meaningful context is extracted by the sophisticated classification and time series analysis for the event-level sequences. The series of event patterns and interpreted knowledge are embedded in the analysis module to recognize abnormal conditions instantly. Moreover, our system opens an interface to define the relationship between the environmental aspect and facility control equipments, set the control action trigger condition, and integrate new event detection logic. The inference engine defines a set of rules to detect events. To define a rule, each sensor and node is assigned a unique identifier, while max(), min(), average(), count(), and run() functions are provided for better event specification. Using this, we can specify several rules, for example, report an event when average temperature of node 123 is higher than 35, or turn on all fans installed in sensor node 452. Based on this rule-base, the middleware checks the validity of the sensor data and requests the retransmission if it has an error term. After calculating the difference from the previous sensor reading, the middleware detects an abnormal condition based on the empirically obtained event patterns and knowledge. This procedure is illustrated in Figure 4.
Intelligent Ubiquitous Sensor Network for Agricultural and Livestock Farms
Data logging mgmt Data management
201
Sensor management Sensor group mgmt dynamic event
Query processing
MSMQ (MicroSoft Message Queue) Data conversion
excel extension, XML exchange
Error processing
exception check, error policy
Duplication processing
duplicate sensor data check
Event processing
event detection logic
Sensor interface Fig. 3. Middleware architecture Rule
Event detection Rule Sensor info Database manager Rule
Data mining
Sensor value Log
Rule−based value processing
Response & event Request
Actuator control process
Data manager
Sensor value
Data source (Sensor nodes)
Fig. 4. Event logic
4
User Interface and Control Action
Remote monitor-and-control keeps track of the environmental sensor values on temperature, humidity, lightness, and CO2 . It can turn on or off the power switch to each actuator device. For example, the temperature monitor tracks the current temperature of a specific position selected via the geographic map. According to the initiation command, the server module begins to collect and store sensor readings. During the lifetime of this operation, the event detection is carried out based on the criteria specified in the query. The client also retrieves the current temperature value to monitor the up-to-date temperature change. Figure 5 show the user interface implemented in this application. First of all, it displays the
202
J. Lee et al.
(a) Normal status
(b) Abnormal status detection Fig. 5. User interface
Fig. 6. Control box interface
map, location of sensors along with the current status of them. In addition, the series of sensor values are scrolled in the listbox, while a graph is created to plot the temperature change. Figure 5(a) indicates the normal status where no sensor value deviates from the given bound, and all nodes are marked blue. Whereas,
Intelligent Ubiquitous Sensor Network for Agricultural and Livestock Farms
203
in Figure 5(b), one sensor node detects the value out of the normal range and this node turns red. In addition, a remote control interface is implemented in an embedded control box and a u-Multi smart mote. First, the embedded control box application interacts with the control system via TCP/IP. As a Microsoft Windows application, it sends a control command such as current status retrieval and specific control action trigger, as shown in Figure 6. In response to this command, the sensor network sends an acknowledgment back to the controller box. In addition, the sensor network can automatically notify the event of approaching the final permissible borderline of current sensor reading. This control box application is also implemented as a web program. Second, u-Multi smart mote is functionally similar to the control box except that its communication interface is SMS (Short Message Service) instead of TCP/IP and it is developed on a smaller user display. One of the most critical events is power breakage and the battery can survive tens of minutes to allow a failure recovery procedure, including notification to human managers.
5
Concluding Remarks
In this paper, we have designed and developed an intelligent ubiquitous sensor network targeting at agricultural and livestock farms which have a variety of sensors and create a great volume of sensor data records during the monitoring phase. For the sake of efficiently and accurately detecting the specific events out of the great number of sensor data which may include erroneous terms, correlative components, the middleware module embeds an empirical event patterns and knowledge description. It also interprets sensor-specific data to the actual values. For the filtered data, data mining module opens an interface to define the relationship between the environmental aspect and facility control equipments, set the control action trigger condition, and integrate new event detection logic. Finally, the remote user interface for monitoring and control USN is implemented in Windows, Web and mobile device applications. As future work, we are planning to design an advanced data inference engine for management information as well as sensor data [13]. The sophisticated data analysis will create a new type of management messages and those messages will make USN more intelligent.
References 1. Golab, L., Oszu, M.: Issues in Data Stream Management. ACM SIGMOD Record 32, 5–14 (2003) 2. Lee, J., Park, G., Kim, H., Kim, C., Kwak, H., Lee, S., Lee, S.: Intelligent Management Message Routing in Ubiquitous Sensor Networks. In: Int’l. Conference on Computational Collective Intelligence-Technologies and Applications (2011) 3. Madden, S., Franklin, M., Hellerstein, J., Hong, W.: TinyDB: an Acquisitional Query Processing System for Sensor Networks. ACM Transactions on Database Systems 30 (2005)
204
J. Lee et al.
4. Madden, S., Franklin, M.: Fjording the Stream: An Architecture for Queries over Streaming Sensor Data. In: Proc. of the 2002 Intl. Conf. on Data Engineering (2002) 5. Ghanem, T., Elmagarmid, A., Larson, P., Aref, W.: Supporting Views in Data Stream Management Systems. ACM Transactions on Database Systems 35(1) (2010) 6. Culler, D., Estrin, D., Srivastava, M.: Overview of Sensor Networks. IEEE Computer 37, 41–49 (2004) 7. Lee, J., Park, G., Kwak, H., Kim, C.: Efficient and Extensible Data Processing Framework in Ubiquitous Sensor Networks. In: International Conference on Intelligent Control Systems Engineering, pp. 324–327 (2011) 8. http://www.tinyos.net 9. Cuevas, A., Urue˜ na, M., Laube, A., Gomez, L.: LWESP: Light-Weight Exterior Sensornet Protocol. In: IEEE International Conference in Computer and Communications (2009) 10. Kang, M., Kang, D., Crago, S., Park, G., Lee, J.: Design and Development of a Run-time Monitor for Multi-Core Architectures in Cloud Computing. Sensors 11, 3595–3610 (2011) 11. Ulrych, J.: Processing XML data streams: A survey. In: WDS Proc. Contributed Papers, pp. 218–223 (2008) 12. Madden, S., Franklin, M., Hellerstein, J., Hong, W.: The Design of an Acquisitional Query Processor for Sensor Networks. In: ACM SINGMOD (2003) 13. Woo, H., Mok, A.: Real-Time Monitoring of Uncertain Data Streams Using Probabilistic Similarity. In: Proc. of IEEE Real-Time Systems Symposium, pp. 288–300 (2007)
